This readme.txt file replicates results in "Subsidies and Structure: The Lasting Impact of the Hill-Burton Program on the Hospital Industry". 

The analysis was conducted using Stata/SE 10.0 for Windows on a PC using Windows 7, using historical data from the American Hospital Association's August issue of "Hospitals." Electronic versions of the hard-copy data were provided by Amy Finkelstein, who acknowledges upported by the National Institute on Aging of the National Institutes of Health under Award Number P01AG005842. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. 

The dataset "aha_1948_2006.dta" provides the compiled AHA data from 1948 to 2006. Supplemental data were added to the AHA data, including the Area Resource Files, which can be accessed at http://ahrf.hrsa.gov/. The "arftemp.dta" data provide the ARF data from 1947 to 2006. The following lists variables from the ARF that were extracted:

pop             population
regioncode      census region code                    
regionname      region name                    
nfmd            number of non-federal medical doctors
POP65           population over 65
employm         total employment            
nwpop           non-white population              
fam             total number of families          
urbpop          urban population      
POPLT5          population less than 5              
popdens         population density        
medfaminc       medican family income

Because many of these variables are availabe from the decennial Census, years with missing values were interpolated using the "ipolate" command in Stata. Those variables begin with "i" (e.g., the interpolated population variable is "ipop", the interpolated non-white population is "inwpop", etc) in the "arftemp.dta" dataset.

The "hbprtemp.dta" dataset is Hill-Burton project Register data that was generously provided in electronic format by Heidi Williams, who acknowledges supposrt from NIA grant P30-AG012810 through the NBER. Those data were collapsed to the county-level by matching FIPS codes to county names to create "hbprtemp.dta".

Lastly, the "contigcounties.dta" dataset was downloaded from ICPSR at http://www.icpsr.umich.edu/icpsrweb/ICPSR/studies/9835. Those data were used to construct the "adjcounty_fund.dta" data, which indicates whether or not a county's adjacent county received Hill-Burton funding.

The data from aha_1948_2006.dta were cleaned and manipulated to create the main working dataset, "data_working.dta", which is used in the following programs to generate results in the paper. The code to generate "data_working.dta" from the AHA panel is entitled "create_data_working1.do".


*	*	*	*	*	*	*	*	*	*	**
* LIST OF RELEVANT CODE TO REPLICATE RESULTS IN PAPER *
*	*	*	*	*	*	*	*	*	*	**

adjspec_49_54controls1.do
- Creates national and tercile regressions for adjacent county spillovers analysis
- Table 3


bytercile1.do:
- Treatment dummy regressions by tercile
- Summary stats by tercile
- Treatment effect on beds pc, by tercile, weighted
	- Without leads
	- With leads
	- Without leads, dropping 9 outlier counties
	- With leads, dropping 9 outlier counties
- Treatment effects on admits pc, by tercile, weighted
	- WIthout leads
- Predicition exercise using treatment dummies by tercile
- Prior regressions by tercile using the ever-treated counties only (dropping control never treated countie)
- Figures 5 and 7


create_adjcounty_fund_data.do
- Creates "adjcounty_fund.dta" using the Contiguous County data


create_data_working1.do
- Create main working data file "data_working.dta" from raw files


distbn1.do
- Summary statistics for distribution effects
- Table 2


fundingpc1.do:
- This has all regression we have using funding pc as the main explanatory variable of interest:
- Funding pc regressions on beds pc, national, weighted
- Funding pc regressions on beds pc, terciles, weighted
- Table 4


HBpaper_natltrends1.rdata:
- National trends of beds and admissions pc
- Figure 3


IV_49_54controls1.do
- Creates IV regressions using priority as instrument
- Table 2


longterm1.do
- Long-run effects of HB funding on beds pc, by ownership type
- Table 6


main1.do:
- This has all the national level regressions using treatment dummies:
- Treatment effects on beds pc, national, weighted
	- Without leads
	- With leads 
	- With leads, dropping 9 outlier counties
- Treatment effects on admits pc, national, weighted
	- Without leads
- Table 1, Figures 4 and 6


prediction1.do
- National and tercile prediction exercise using funding pc 
- Table 5



